Techniques of allowing a robot to converse with a user for the purpose of communication with the user are known. In most of such techniques, conversations are carried out by selecting an operation pattern of a robot based on speech recognition of speech uttered by a user and causing the robot to utter speech or to move hands and feet based on the selected operation pattern.
A technique for such speech recognition is disclosed in Japanese Unexamined Patent Application Publication No. 2008-107673. This relates to a technique of allowing a conversation robot to analyze speech uttered by a user, calculate an emotion parameter, and perform speech output and the like based on a scenario of speech, a look, or a motion determined depending on the calculated emotion parameter.
However, in the technique disclosed in JP '673, since the emotion parameter is calculated by calculating a parameter from appearance frequencies of minute high frequencies and low frequencies in the speech input from the user, there is a problem in that it is difficult to uniformly cope with a variation in expression for each emotion of a user and a user emotion recognition rate is low.
Therefore, it could be helpful to control operation of a robot in a conversation by exchanging messages to improve a user emotion recognition rate in the conversation between the user and the robot.